In the trust literature, there are many research work that build models to evaluate trust. The trust models range from simple models, such as eBay, AllExperts, Epinions, and Amazon models, to complicated models [86, 100, 25]; they cover trust [46, 45, 103, 50], reputation [52, 100, 57, 89], or hybrid approach [38, 103]. While some models consider some trust aspects [52, 46, 45, 103, 41, 38, 50], most do not consider trust bootstrapping [46, 45, 103, 41, 38, 50, 86, 100]. Thus, a trust model should be simple, understandable, comprehensive in covering different trust
aspects, and extendible. In addition, it should consider trust bootstrapping and resolve various trust challenges.
Malik and Bouguettaya [52] adopt a model for reputation bootstrapping, however, their model does not consider different trust aspects or address trust challenges such as cold start and whitewashing. Kim [45] and Kim and Doh [46] build a model to select the optimal services composition execution path based on trust and QoS. The authors assume that a trust type is associated with each service, where the assignment of trust types is performed by the client or by a trust authority ( i.e. T T(si) = a, the assignment of a trust type ‘a’ to a component
Web Service si). In this model, the execution plan with the maximum value of profit is the
optimal one, where the profit is a function of trust weight and quality vector MQ, computed QoS attributes, of the execution plan. However, the model is based on undefined trust information and assumed trust types, and it does not consider bootstrapping.
Zhengping et al. [103] build a trust model where trust degrees are based on the properties of services and recommenders. In this model, there are known and unknown recommenders based on whether they are familiar to the trusted authority. The trust degree of a service is related to the service descriptionVkp, known party’s recommendationVup, and unknown party’s
recommendationVa. Thus,trustdegree(A, S) = f(Vkp, Vup, Va). However, their model does
not involve many trust aspects, such as the context-specific property of trust, and it does not consider trust bootstrapping.
Kalepu [41] calculates the compliance values for all of the services SLA parameters, which involves the difference between the projected and achieved parameter values. The result is expressed as positive, negative, or zero compliance. A positive compliance indicates that the agreed-upon values have been delivered without violations, while a negative compliance means that the provider failed to deliver the agreed-upon values. Finally, a zero compliance indicates an ideal value, where the delivered values are exactly equal to the agreed-upon values. The local and global rankings of the services and their providers are evaluated based on the com- pliance levels of SLA parameters. However, this model does not consider trust bootstrapping. Maximilien and Singh [59] derive a trust function based on QoS, where some QoS parameters are preferred by service customers if their values are high, such as availability, and others are preferred if their values are low, such as response time.
Jin-Dian et al. [38] propose a WSTrust model based on feedback and reputation, where requestors provide their feedback to the service broker about the services they have consumed. This trust model considers the following trust aspects and challenges: trust and reputation vary when the number of interactions increase, trust is based on direct experiences and recommenda- tions, the prevention and punishment of repeated malicious behaviours by users, the addressing of users’ unfair ratings, and trust is context-dependent and subjective, which involves dividing trust into three trust relationships: brokers trust services, requestors trust brokers, and requestors trust services. However, the authors cover a limited number of trust aspects and do not consider trust bootstrapping and other trust challenges. Similarly, Liu et al. [50] build a trust model that only considers three trust properties: specific, a matter of degree, and dynamic. The literature contains other complicated trust models that consider one or more trust aspects [86, 100, 25]. However, trust models should be simple, understandable, and comprehensive.
Some research work use the monitoringapproach to check QoS compliance, ensure SLA, and rank services based on QoS attributes [83, 103, 101, 41, 90, 87, 69]. Rosenberg et al. [83] propose an evaluation approach for bootstrapping QoS attributes of Web Services that provide a set of up-to-date QoS attributes for Web service selection. Zhengping et al. [103] monitor the behaviour of services at run time, and Zhang et al. [101] present a Web Service search engine to find desired Web Services. Specifically, the engine ranks Web Services by monitoring the non-functional QoS characteristics of Web Services. Kalepu et al. [41] use a third party to monitor the transactions between requestors and services for detecting any SLA violations. The third party verifies the values of SLA parameters in the agreement against the obtained values by probing or intercepting the client invocation. Furthermore, Vu et al. [90] monitor QoS for Web Service ranking and selection and to detect and address false ratings. Sherchan et al. [87] measure the compliance of QoS attributes by comparing the projected values agreed-upon in the SLA and the delivered values obtained from the performance monitoring system. Nguyen et al. [69] build a trust and reputation model for Web Services based on feedback and QoS monitoring approaches.